Monthly active users from an event log

Question

An event log has one row per session with a `user_id` and an `event_time` string. Product wants monthly active users — the count of **distinct** users per calendar month.

Write a function `mau(df)` that returns a Series indexed by the month as `"YYYY-MM"` strings, holding the distinct user count, s

Accepted Answer

Step 1: `pd.to_datetime` turns the strings into real timestamps — string slicing would work here but breaks the moment the format changes. Step 2: `.dt.strftime("%Y-%m")` produces the `YYYY-MM` bucket key. `.dt.to_period("M")` is the alternative when you want a Period index rather than strings. Step 3: Grouping by that Series (not a column name) is legal in pandas as long as its index aligns with the frame. Step 4: `.nunique()` counts *distinct* users, so user 2's three February sessions count once. Using `.count()` here is the classic mistake — that returns sessions, not users. Step 5: `.sort_index()` puts the months in chronological order, which zero-padded `YYYY-MM` strings sort correctly.

Monthly active users from an event log

More Dates & Time Series questions